The improvement of a software design methodology by encapsulating knowledge from code

While numerous software engineering methodologies exist, are well documented, and are widely used, e.g. JSD, SSADM, etc., it can be argued that none cater for the specific problems associated with reengineering legacy, or dusty-deck systems. This paper reports on a development environment, KATT, that enshrines a methodology which enables the development of new codes or the reengineering of existing legacy codes. The system is designed to enable sequential codes to be translated for execution on either parallel or sequential architectures, e.g. moving Fortran 77 to a workstation cluster or translating Cobol to Java. The approach involves the encapsulation and application of restructuring knowledge via the use of expert systems and neural networks.